(1'9'6'8'""159-83 and in Later Publica- (1972 539

Home , Prothesis (linguistics), Rhotacism

FINAL WEAKENING AND RELATED PHENOMENA1 Hans Henrich Hock University of Illinois at Urbana-Champaign

1: Final devoicing (FD) 1.1. In generative phonology, it is a generally accepted doctrine that, since word-final devoicing (WFD) is a very common and natural phenomenon, the ob- verse phenomenon, namely word-final voicin~ should not be found in natural language. Compare for instance Postal 1968 184 ('in the context----'~ the rules always devoice rather than voice'), Stampe 1969 443-5 (final devoicing comes about as the result of a failure t~ suppress the (innate) process of final devoicing), Vennemann 1972 240-1 (final voicing, defined as a process increasing the complexity of affected segment~ 'does not occur.')o 1.2 One of the standard examples for WFD is that of German, cf. Bund Bunde [bUnt] [bUndeJ. However Vennemann (1'9'6'8'""159-83 and in later publica- tions) and, following him, Hooper (1972 539) and Hyman (1975 142) have convincingly demonstrated that in Ger- man, this process applies not only word-finally, but also syllable-finally, as in radle [ra·t$le]3 'go by bike' (in some varieties of German). The standard view thus must be modified so as to recognize at least one other process, namely syllable-final devoicing (SFD). (For a different eA""Planation of this phenomenon compare section 2.3 below.) 2· Final voicing (or tenseness neutralization) 2 1 A more important argument against the standard view, however, is that, as anyone with any train- ing in Indo-European linguistics can readily tell, there is at least one 5roup, namely Italic, where there is evidence for the allegedly impossible final voicing, cf PIE *siyet > OLat. sied 'would be'.

219 220 1975 MID-AMERICA LINGUISTICS CONFERENCE

2 2 According to Miller (1973 692), this is an example of 'an 11 unnatural 11 or 11 less natural" rule', FD being more natural--although the most 'natural thing is for nothing to happen in final position (as in English, French, etc.). There is no reason for a boundary to condition either voicing or devoicing The presence of either of these rules will be motivated by language-particular pause and boundary conditions.' At pauses, of course, 'the vocal cords must be open which vrill predict devoicing, if pause is anticipated ' ( fn 2, p 711) • Though Miller is, of course, correct in stating that nothing may happen at all, this is not a cogent argument against the claim that FD (or final voicing) may be a natural development For this claim evidently addresses its elf merely to the question '\vhat happens most naturally word-finally (or syllable-finally), if there is any change at all?' On the other hand, Inller' s argument that word ooundary is not a plausible phonetic environment for final (de)voic1ng is well taken, as is his claim thao pre~ position is a plausible phonetic environment, namely for (anticipatory) devoicing ( assimilatJon). As I have argued elsewhere (Hock, In press §5.1 w. fn 14), word-final devoicing, then, is to be explained as the result of (rule) generalization. However, this still \lould not account for final voicing 2 3 To my knowledge, it was Andersen (1969a,b) who first proposed an explanation of final voicing (not in Italic, buo in early Slavic) within the generative phonological paradigm. According to Andersen, the change of the preposition *ot to *od in dialectal Com- mon Slavic was the result, not of the usually recogrnzed process of final voice neutralization, but of final tenseness neutralization Just as in final voice neutralization it is the unmarked value of the feature voice], namely [- voice], which appears in the unmarked word-final position, so in final tenseness neu- tralizavion, the unmarked value of [+ appears, namely [-tense]. And it was the feature tense] FINAL WEAKENING AND RELATED PHENOMENA 2~1

which was 'phonemic' in early Slavic This 1s shown, in Andersen's vieu, by the fact that there was lenition (of~>£, > dz>!) in early (dialectal) Slavic. This explanation, though effective for early Slavic *at > *ad, hmJever, evidently causes problems in the case of-a language like German where [+ tense] is 'phonemic 1, 4 but where it is [+ tense] which appears in the word- or syllable-final position of neutralization (cf 1.2 above). To account for this phenomenon, Jmdersen (1972 44-5) assumes that unlike word-final position, the syllable-final position is 'marked', and that in this 'marked' environment it is the marked value of the feature [±tense], i.e. [+ tense], \Jhich appears. 2.4: This explanation of German SFD was accepted also by Anttila (1972:199-200) who argues that the syllable learning process in child language acquisrtion (first CV, then CVC, then VC, and finally V) shows that syllable-final environment i5 indeed 'marked in respect to syllable-initial position.' Moreover, he argues that 'Voicing i5 learned earlier in syllable-initial position, for example, \n1en an English-speaking child can already produce [buk] 'book' he goes on saying [pik] I pig I • • • I 2.5 A further boost for the hypo~hesis that 11 11 final voicing , i.e final laxing may be a genuine 1 alternative to FD came with Szemerenyi's paper of 1973. Szemerenyi first established that ~he final voicing of Italic has direct counterparts in Germanic and Celtic. I1oreover, he showed that there are early Iranian and Slavic counterparts to the rather unusual external voicing sandhi of Sanskrit [+stop]--> [c:it.voice] I __ # [~voice] with voicing sandhi across word boundary applying both before [M voice] and before [U voice] segments, while internally, the more natural assimilation [ + stop J --> [ voice] I [+ oostr. J °' -- °'voice is found, with voicing sandhi applying only before [r1 voice] segments. Thus, externally we find Skt 222 1975 MID-AMERICA LINGUISTICS CONFERENCE

/vak#asya/ realized as vagasya 'his speech' vs. internal /vak-as/ --> vacas 'of speech'. It is only before obstruent that external and internal sandhi agree, as in /vak-bhis/ --> vagbhis 'by speeches' = /vak#dhiyate/ --> vagdhiyate 'speech is meditated (upon)'. Apparently without being aware of Andersen's (1969 a,b) similar argument for early Slavic, Szemerenyi then proposed to explain both of these sets of 'unusual' phenomena by assuming that in Proto-Indo-European, final obstruents were neutralized in terms of the feature [±tense], in favor of the unmarked value, [-tense]. Italic, Celtic, and Germanic would then retain the resulting final [- tense] obstruents, while in Sanskrit, early Iranian, and early Slavic there must have been a generalization of the internal sandhi rule [+ stop] --> [~tense] I c: tobstr.] -- 111o ense to external sandmconditions. Note that as a consequence of this analysis, Sanskrit winds up \Ti th a more natural and 'usual' 'voicing assimilation' in external sandhi than under the traditional analysis in terms of the feature [±voice]. , 2.6 Though synchronically, Andersen's and Szeme- renyi's analysis may well be correct, I am not con- vinced that it must be correct also historically 2.6.1 First of all, as I hope to have shown elsewhere (Hock, In press.~5.1), it is possible to account for the external voicing sand.hi of Sanskrit (as well as of early Iranian and Slavic) as the result of an analogical generalization of final devoicing, reinterpreted as voice neutralization. Such a neutralization can then oe manif estea through the rule [+ stop] --> [~ voice] I # [OC voice] by overgeneralization of the internal voicing rule [+ stop] --> [«voice] I [+ obstr.] -- OC. voice • 'As a generalization (of voice neuLJralization before all(') instances of word boundary), this process is not bound by the same conditions of phonetic natural- FINAL WEAKENING AND RELATED PHENOMENA 223

ness as a purely phonetically conditioned change.' There is ample evidence supporting this assumption, both in terms of other sandhi (over)general1zat1ons in Sanskrit, and in terms of parallel developments in Polish dialects (cf Vaillant 1950 101 and 209) It is interesting to note that Andersen (1969b 302-3 w. earlier literature) provides conclusive evidence for a similar generalization in Slovak, as in (*su vod(o~)9 >) s vodou [z vodou] 'with water'. Relics of an earlier stage, in which there was no such overgeneralization of voice assimilation are found in ex- pressions containing personal or anaphoric pronouns, as ins nam1 [s nami] 'with us'. It is then possible to argue that this overgeneralization may have occurred in Froto-Indo-European and that, because of the resulting occurrence of voiced stops in most environments, it was possible to take these voiced stops as the basic word-final manifestation of stops, and- to generalize~ them, presumably reinterpreted as [- tense] stops, to all environments. This uould account for the Italic, Celtic, and Germanic situation .. Note that this explanation thus removes the difficulties of Brugmann (1897 883-4, 885-6) who uas in- clined to view the Sanskrit external sandhi pattern as analogical,) but felt that in light of Italic *-t > -d, it may have been inheritedfrom Proto-Indo-European-;-6 thus presumably precluding an analogical origin. Moreover, the analysis here proposed as a possible alternative has the advantage that, unlike that of An- dersen and Szemerenyi, it offers a phonetically motivated starting point for the phenomena under discussion~ namely (prepausal) FD. (But cf also section 3 below.; Finally, it might be argued that the evidence of the extant Sanskrit texts, where the prepausal form of final obstruents invariably is voiceless (as in /pad##/ --> p@t 'foot') would be more compatible with an analysis w ich postulates voicing, rather than tenseness neutralization. However, as the subsequent discussion 224 1975 MID-AMERICA LINGUISTICS CONFERENCE

in this paper will show, the Sanskrit facts are open to various interpretations. 2.6.2 There are difficulties also with Ander- sens attempt to explain German SFD as natural in a language with 11 phonemic" [.±.tense]. First of all, German shows this neutralization also word-finally, i.e. in 'unmarked' environment, contrary to Andersen's (and Szemerenyi's) analysis which would predict [- tense] for this environment Notice that the supportive evidence from child language acquisition adduced by Anttila, at closer investigation, turns out to be equally damaging as the evidence from German For Anttila's examples of monosyllabic words show that [+ tense] (or [- voice]) is natural not only syllable- finally, but, again, also word-finally.

In addition, note that there are languages where both [- tense] and [+ tense] obstruents can appear as the realization of neutralized final obstruents. Thus, Danish offers a free variation of voiceless tense and voiceless lax stops, both for final voiced stops and for final voiceless stops, as in tab [-b/p] and tak [-g/k], cf. Arnholtz and Reinhold-r9"36 ~9-60 b ~si miiar situation may have obtained in some varieties of Sanskrit, for Pa~ini (8.4.56, cf 8 2 39, 8 4 53-55) allows both voiceless and voiced final stops before pause. But, as pointed out earlier, the Sanskrit facts are open to various interpretations.) Should we claim that languages of this sort have both "phonemic" tense] and 11 phonemic 11 [.±.voice]?! 2.6.3 Also Andersen's argument that the occurrence of lenition in early Slavic ind:u:ates that Slavic had 11 phonemic 11 [+ tense] is open to doubt. For as Old Enghshj with meuial voicing (as in PGmc. *\lulfos/z > [nulvas 'wolves') and final devoicing (as in pre-OE hla1(l > OE hlaf 'loaf') shows, 11 lenition11 or--as I pre- fer to call it--medial weakening (for uhich cf. sec11ion 4- 1-2 oelm1) may \Tell occur in languages with 1lFD, i.e in languages which according to Andersen's interpretation should be considered to have "phonemic 11 voicG]. FINAL WEAKENING AND RELATED PHENOMENA 225

3: Final weakening (Fw) Moreover, there is reason to believe that final voicing or laxing--where it is a genuine, rather than secondary development--may conceptually belong to a larger, more general set of possible final developments, namely final weakening This would indicate tnat the concept of final (tenseness) neutralization merely describes the results of a historical cnange, rather than (the motivation of) the change itself. 3.1 Though languages with word-final weakening (WFW) are probably not as numerous as languages with \JFD, an impressive array of languages with WFw can be found.9 They can be conveniently arranged under the following headings. 3.1.1: Word-final fricativization: Italic (or Osco-Umbrian) t > s I n # (Gauthiot 1913:135; von Planta 189'2':58'2') - Old French .! > g I (Ewert 1943 74-5, 76-~ Old Irish £ > .£ I ___JI. (Pedersen 1909 430) Cornish .! > ~ I ___JI. (ibid.499-500) Cf. also 3.1 below. 3.1.2: Word-final voicing· Old Irish ~ > ~ I ____j (Pedersen 1909 133) Cf. also section 2 above(?). 3.1.2: \Jord-final gliding. Italian s > i I V # (via ) (Grandgent 1927:76-7) - - pre-Sanskrit (s >) h > z I #[+ voice] (Allen 1962:71 and 101-E) ..L -

3 .1.4: ~'lord-final rhotacism Elean Greek ~ > £ I ___JI. (Schwyzer 1939 410) 11 New 11 Umbrian ~ > E I __# (von Planta 1892 226 1975 MID-AMERICA LINGUISTICS CONFERENCE

582) 3 1.5 Word-final flapping: 01d Latin (*t/d >) ~ > .:£ I _#C~s1!~·J (Safarewicz 1969 102) 3.1.6. \lord-final feature depletion Eastern Arabic t > h I ___jf. (Gauthiot 1913 17) (via -.§. ?) -- dialectal Oscan d > h I _# (von Planta 1892 577-8, Buck 1904 84) -- Finnish t > ? I ____jflO (Gauthiot 1913 102-6) -- 10 Finnish ~ > :!_ I __}f (ibid ) Chinese dialects [+ stop] > :!_ I (Chen 1973) Sanskrit s > ~ / # (Gauthiot 1913:113-21, Allen 1962. 71, 101=8) - Spanish dialects s > h I # (Alarcos 1961 271, I;lBlmberg 1948 393-7) --- 3 1.7 Word-final loss Class Lat. (*t/d >) d > ¢IV # (Safarewicz 1969 102) -- - Arabic dialects ! > ¢ I __JI. (Gauthiot 1913 l~ Finnish (t,k >) :!_ > ¢ / ___# (ibid 102-6) Chinese dialects([+ stop]>) ? >¢I # (Chen 1973) -- Italian~> ¢ I [_ ~cct.J ___ff (Grandgent 1927 76-7) later OFrench 2 > ¢ I _# (Ewert 1943 74-7) Spanish dialects (s >) h > ¢ I # (Alarcos 1961 271, Malmberg 1948 393=7) --

~his change is, of course, widely attested in many o"'Gner languages FINAL WEAKENING AND RELATED PHENOMENA 227

3 2. The same set of phenomena can also be found in syllable-final environment. That is, there is evidence for syllable-final weakening (SFW) 11 3.2.1: Syllable-final fricativization This process is found in \Jestern Romance (Alarcos 1961 234, Ewert 1943 71, Fouche 1927 59) and again, more recently, in Spanish (Alarcos 1961 178-9, Hooper 1973 70-102, Malmberg 1948 393-7 and 1952.409-12, Nava- rro Tomas 1918.61-79). Moreover, it is found in Greek (Martinet 1970.335, Schwyzer 1939 211)_, Celtic (Peder- sen 1909:75, 93, 123, 429-30, Thurneysen 1961 135, 136, 139, 140), and in Iranian (Reichelt 1909 38 and passim, cf. also the discussion in 7.2.5 below). 3.2.3· Syllable-final voicing This process is found, combined 11ith fricativization, in Spanish (cf. the references in 3.2.1). 3 2 3 This process is found, as a continuation of fricativization, in Western Romance (cf. 3 2.1 above), in Spanish dialects (Hooper 1973 70-102, Malmberg 1948 393-7 and 1952.409-12), and in British Celtic (Peder- sen 1909 75, 93, 123, 429-30). 3 2 4. Syllable-final rnotacism

Compare dialectal Spanish desde > derde (~larcos 1961 271) Syllable-final flapping This is perhaps found in OLat (ad- >) arvorsum, arfuisse , if the univerbation of prefi'x and verb stem preceded Old Latin flapping. 3.2 6 Syllable-final feature depletion This is found in Spanish dialects with s > h (Alar- cos 1961 271, Malmberg 1948.393-7 and 1952 409-12) and in Oscan-Umbrian (*~ >) *~ > ~ (Diver 1953. chapter 5, 228 1975 MID-AMERICA LINGUISTICS CONFERENCE

Martinet 1970:337). Considering that in Finnish, word- final ? combines with following [- syll.] segments into geminates, it is possible to include in this category also developments like Lat. octo > It. otto. 3.2.7 Syllable-final loss (with or without com- pensatory lengthening)· This change is found in later Oscan-Umbrian h > 0 (cf. 3.2.6), as part of the Slavic open-syllable conspiracy (Martinet 1970 349, cf. also 7.4 4.1 below), and in innumerable other languages. 3.3.1 Terms like 'weakening' are frequently em- ployed in traditional historical linguistic publicatJOm in reference to phenomena like those above. Thus, Buck (1904 84) refers to the dialectal Oscan word-final change of d to h as a 'ueakening', Alarcos ( 1961) refers to the \/estern Romance shift of Lat kt, ks to pre-Span. yt, ys as a 'tendencia • debilitadora'(.2}4), and to the change of s > h in dialectal Spanish as a 'debilita- miento', and Cnen tl973) refers to the common development of final stops in Chinese dialects of the type t > !, > _: > 0 as 'feature depletion' • - 3.3 2 Similarly, Allen (1953 70 and 1962 98) argues that the word-final stops of Sanskrit have a 'lax' or weak articulation (of the type [~, ? > 0 or k > x> h > y > 0, thus establishing that-they are in- timately-related to each other, seems to require 3 FINAL WEAKENING AND RELATED PHENOMENA 229

general coherent theory. 3 3 3.1· The first such theory known to me is that of Gauthiot (1913). Accepting de Saussure's theory that syllable-final (or preconsonantal, prepausal) 11 11 obstruents are (naturally) implosive , i.e. unreleased (1889, cf also the posthumously published 1916.79-93), Gauthiot proposed that word-final weakening, of the type t > ? > ¢, t > d, t > s, s > h, is the normal development-of final 11 imp~osives-rr. - 3.3.3.2: This theory was further extenaed and motivated by I•1almberg ( 1948 393, 396, and 1963 68) As already de Saussure (1916 79-93 with 71-6) had postulated, the syllable is a sequence of explosive segments of increasine; aperture, followed by a sequence of implosive segments of decreasing aperture, 1vhere the de- grees of aperture are defined as follows. (1) ¢-aperture stops (2) aperture 1 fricatives (3) aperture 2 nasals (4) aperture 3 liquids (5) aperture 4. .2:., ~' u (including their voiceless counterpart, h) (6) aperture 5~ ~' £, £ (including their voiceless counterpart, h) (7) aperture 6- ~ (including its voiceless counterpart, ~) A similar ranking of 'sonority' had been indepen- dently proposed by Jespersen (1904 186), although Jes- , persen's discussion of the concept of the syllable (185-203) betrays a much more agnostic attitude concerning the possibility of adequately defining it Still, Pedersen did propose that between a given segment and the peak of the syllable (which definable), only segments of the same, or of a higher sonority are tolerated. (19l)o (1) voiceless (a) stops (b) fricatives (2) voiced stops 230 1975 MID-AMERICA LINGUISTICS CONFERENCE

(3) voiced fricatives (4) voiced (a) nasals (b) laterals (5) voiced r-sounds (6) voiced high vowels (7) voiced mid vowels (8) voiced low vowels These two theories were combined and further modified by Grammont (1939 38,39,43-5,98-100). According to Grammont, we must distinguish between explosives and (syllable- or word-final) implosives. In explosives, muscular tension increases, they are therefore referred to as 'increasing' ('croissant:!). Implosives, in which muscular tension decreases, are referred to as 'decreasing' ('decroissants') In addition, there is an aperture hierarchy of the follouing sort. '\ ¢ stops 1 spirants 2 nasals 3 liquids 4 semivowels 5 high vowels (including v and h) 6 mid vowels ( II "ff 11 ,,.) 7 low'vowels ( 11 11 II ll) The syllable is then defined as 'a sequence of increasing apertures followed by a sequence of decreasing apertures' ( 99). Syllable-initial stops are explosive and 'increasing'~ syllable-final stops, implosive and 'decreasing' (98;. Following this (refined) theory of Gramrnont's, and on the basis of a rich collection of Spanish data illus- trating syllable- and word-final weakening, Malmberg (1948) proposed the theory that 'La place apres le support syllabique est plus ufaible 11 qu' ailleurs. Il est done normal que la distinction reguliere et consciente des differences phoniques soit realisee plus difficilement qu'a l'initiale de la syllabe ou la force articulatoire est concentree.'(~93) 'C' la evidemment un phenomene de phonetique Cet affaiblisse- FINAL WEAKENING AND RELATED PHENOMENA 231

ment de la partie finale est propre [comme ten- dance] a chacque syllabe independamment de la langue.'(396). Similar arguments can be found in Malmberg 1963 46-9. ~lhile the concepts of de Saussure, Jespersen, and Grarrunont may not al\Jays be applicable synchronically, they do show tneir validity in historical change Thus, Hhile the frequently occurrins #st violates Jes- persen' s theory concerning the sequence-or sonorities, histoncal changes, such as Lat #st > 1st/est in Romance, confirm the ultimate validity of this theory Similarly, tne common \Jeakening or loss of final implosives illustrates the essential correctness of tne theories of de Saussure and Grammont Malmberg's theory has subsequently been approved py Pulgram (1970 74-5) and, in (natural) generative phonology, by Hooper (1973 90-102). 3.3.3 3 \lhile Gauthiot's and Malmberg's theories go a long way touard providing a general, coherent theory of final weakening (in terms of syllable struc- ~ure), they are not without their weaknesses First of all, \1hat is left undefined and vague is the concept of weakening itself. As the presentation in sections 3.1 and 3.2 has shown, there are at least seven different processes sub- sumed under this notion, namely fricativization, voicin~ gliding, rhotacism, flapping, feature depletion, and loss. As mentioned earlier, the fact that (many of) these processes 'feed' each other (in the same, constant environment) suggests that there is some strong, siGni- ficant relationship between these processes This impression is reinforced by ~ne fact that, except for the occurrence of final devoicing, these processes are transitive and irreversible: Examples of reverse shifts (of, say, fricative to stop) are not normally fou..~d in this environment.14 232 1975 MID-AMERICA LINGUISTICS CONFERENCE

rloreover, it ~ possible to account for most of the weakening developments, except loss, in a principled, general fashion by postulating the following (ten- tative) combined hierarchy of aperture and sonority. (1) voiceless stops (2) voiceless fricatives (higher aperture than (1)) and voiced stops (higher sonority than (1)) (3) voiced fricatives (higher aperture and sonority than (1)), voiceless glides including g (higher aperture than (2) and than voiced fricatives, but sonority of (1)), flaps (shorter reduced aperture than ( 1) or voiced stops, and higher sonority than ( 1 )) , glottal stop (maA~mal oral aperture, but sonority of (1)) -- (4) liquids (higher aperture than (1), (2), or voiced fricatives and flaps, and higher sonority than voiceless glides and glottal stop) (5) voiced glides (higher aperture than (1), (2), fricatives, flaps, (4), and higher sonority than voiceless glides and glottal stop)

1 1 It can the.a be argued that 1iJeakening ( ~ther than loss) consists of a 'downgrading' of articulation along this combined sonority/aperture scale. (Note that the omission of the nasals in this scale is motivated nou only by the fact that nasals do not seem to figure in the 1i1eakenings rliscussed .15 It is ~otivated also phonetically, namely by the fact that nasals are [+nasal] stops--i.e in this respect high up in the hierarcny--, uhose relative sonority is a consequence not of oral resonance, as 5n the other segments, but rather of nasal resonance.l )

3.3.3 4 Perhaps even more inportant is the fact that uhere is reason to l:Elieve that the notion of 'implosive' articulation of final obstruents does not provide for a suf.ficiently motivated ultima ratio of final FINAL WEAKENING AND RELATED PHENOMENA 233

,1eaken1ng. For if 'implosive' is defined as unreleased, it is hard to see how this term has any relevance for the weakening of friction continuants like ~ Even more important, even stops may well be re- leased in final position. This is true at least optionally for many varieties of English and seems to be the rule for many varieties of German (at least in careful speech). This suggests that 'implosion' or lack of release, rather than being a universal phenomenon and thus "Ghe possible cause of final weakening (in stops), simply is the first sten of such a final ueakening pro- ~ ess. This is'I'iO'""doU'b"t""what is meant by the notation ! > -£. > 2. > ¢ found in Chen 1973 and similarly in Hiran- da 1974-·53. That is, JUSt like the concept of final (tenseness) neutralization, the concept of final 'implosion' of stops merely describes the result of a change, not the (motivation of the) change itself. 4-. Final weakening and medial weakening ('lenition') 4.1: In its phonetic effects and results, final 't1eakem.. ng appears to be identical with the very common process of medial weakening (lv1r'1) often referred to as lenition. The same types of processes can happen in !'.iW as in FU. 4-.1.1: Medial fricativization. This process has been observed in the history of French (Ewert 194-3.74--7), Spanish (in voiced stops, Malmberg 1952), Tuscan (Izzo 1972 passim, esp.1'73-5), Irish (Thurneysen 1961.74-, Pedersen 1909 4-27-30), Bri- tish Celtic (in voiced stops, Pedersen 1909 ibid ), Younger Avestan (in voiced stops, Reichelt 1909 passim~ Hebrei1 (Martinet 1970 270 w. references).

4 .1.2. I11ed1al voicing This process, wnich frequently may apply to the outcome of medial fricativization, is found in the history of Latin (r•1artinet 1970 335-69), French (E 1ert 234 1975 MID-AMERICA LINGUISTICS CONFERENCE

1943 74-7), Spanish (Malmberg 1952), Italian dialects Grandgent 1927 81-2, 99, cf. also Lausberg 1967.passim), British Celtic (in voiceless stops, Pedersen 1909·427- 30), American English dialects (as in latter [l~dar]), fhddle Indo-Aryan dialects (in voiced stops, Bloch 1934 77-8), cf. also 2 6.3 above 4 1.3 The (*voiceless and *voiced>) voiced stops of rnddle Indo-Aryan may appear as y or v in some- of the later fhddle Indo-Aryan dialects-(Bloch 1934 77-8). 4 1 4 Medial rhotacism Compare Lat. ( *'s >) z > r (IIartinet 1970 335-69), similarly, the outcome *_t!-of Verner's Law (generally) becomes .E in 11/est and North Germanic.

4 1.5 Medial flapping [lrefar]. 4.1 6 Medial feature depletion Cf. English dialectal bottle [ba?+J and note that in the Irish 'lenition' process, s becomes h (> 0) (Thurneysen 1961•74 and passim). - - 4.1.7. Medial loss: This change is found in the historical development of French (Ewert 1943·74-7), Spanish (Malmberg 1952), Irish (cf. 4 .1.6), Middle Indo-Aryan dialects (Bloch 1934 77-8). 4 2 In addition, the relationship between these various processes is analogous to that between the various varieties of FW (for \fuich cf. 3.3.3.3). That is, it is transitive and irreversible. I•1oreover, many languages or language groups which have F\J also have M1v, cf. Romance, Celtic, Iranian (Youn5er Avestan).17 FINAL WEAKENING AND RELATED PHENOMENA 235

4.3: As a consequence it is not surprising that some linguists have--implicitly or eA"':plicitly--considered F1 J and l'LW to be a single, unitary process

Thus, Pedersen (1909.427-30) refers to Celtic 111.1 and FW as lenition hartinet (1970 335-69) subsurries both types of changes in his general discussion of weakening processes. Similarly, Ilalmberg (1952) dis- cusses both sets of phenomena side by side. And ~nder sen (1969a 167-9) considered the evidence of early Slavic lenition probative for his claim tnat early Slavic had final tenseness neutralization, cf 2 3 aoove. Houever, only Hyman (1975 168-9) seems to have at- ~empted to provide a general theory concerning tne ;:i.dentity of the two processes, as well as the reason 1md condition for this identity. Defining FW as final ponsonant depletion, loss, and devoicing ( 'most common'), Hyman continues to argue that M;J occurs intervocalicaIW ~nd FW word- and syllable-finally, the common property of both of these positions being postvocalic This, 'jJhen, must in his vieu be the 'position of weakness ' 4.4: Although Hyman's explanation is interestins and illumina~ive in some respects (cf 5 2.2 belo1), it cannot be considered acceptable.

First of all, if FD is to be included in F1l (but cf. section 5 belou), F ii and M\J cease to be completely parallel (or identical) processes. For devoicing is certainly not a natural development in Medial voiced environi11ent.

lioreover, contrary to Hyman's belief, f'i•l may taY.:e place not only in intervocalic environment, but in any voiced medial environment (except, usually, after nasals), cf. Span. guardar [gr.1arcar] 'guard' (but andar [andar] 'go'). And in such environments it (normally) applies to syllable-initial segments, i e in a position which is clearly opposed to the syllable- final environment for FW. (

236 1975 MID-AMERICA LINGUISTICS CONFERENCE

Perhaps ~ore important than these general considerations, houever, is the fact that even in languages \lhich have both M.\J and FW, the two processes may not have identical manifestations vis-a-vis given segments. Thus, in the prehistory of Old Irish, SFW chane;ed PIE *.12. to b before liquids, and to *P before nasals. On the otner hand, 111.l changed (voiceless) stops to (voiceless) fricatives, cf. Thurneysen 1961 74,139,140. In addition, the fact that ( *ur, *pl >) *br, *bl wind up as 'lenited' ~' e.>.. in Oldlrish(ibidl39)shows that SF1'/ and li'.J \'lere two chronologically different processes, 11itn the latter applying later, and to the 'outpu"G' of the former.18 Finally, note that there are languages uhich have one of these processes uithout the other Thus, Old Ene;hsn has M\i/ (of fricatives), but 1,JFD, not •'lF~J, cf. 2 .6 .3 above. Eean Greek has \lF\'/ of s to r, but no 1-1'.l Similarly, Oscan-Umbrian shous SF~/ (of stops to fricatives), but seems to have no evidence for 'f-'iT1/ (except for ~ > ~ in some of the dialects) 4 5 It can thus be concluded that while--from a General point of view--FW and M1/ are phonetically identical, the evidence of their specific applications suggests that are concepi:;ually different, as uell as conditioned by different envirorunents. 5 Final devoicing and final 'Jeakening 5.1 As mentioned in 4.3 above, Hyman (1975 169) considered FD a subvariety of F.J. Similar claims have been made by Fouche (1927 62) and Schane (1972 210-1) These claims, however, are subJect to several obJections 5.2.1 First of all, FD (and as a result, final voicing) \1ould be the only change 1 1hich is not in con- forni ty with the transitive, irreversible pattern of the F 1.T changes noted in 3 393.3 above IJ10reover, although il'.l is not functionally identical, it is phonetically identical uith \lhat normally has been called FH. If there \Jere in fact a devoicing varJfty of F\v, one shcu.ld expec-c; to find a similar variety of J:vH. Houever., as FINAL WEAKENING AND RELATED PHENOMENA 237

noted in 4.4 above, such a change does not seem to be a natural development in medial voiced environnent 5.2.2 In addition, it might be argued (as ad- umbrated in Hyman 1972) that FU and I U, except for the subvariety of loss, can be considered assimilatory developments, namely assimilations to the more sonorous or 'open' nature of the surrounding seGments--in the case of H1v--, or of the preceding segments--in the case of F'.v. On the other hand, 1tJFD--1r.ihile also an assimilatory process--would differ from HW and Ftv oy being an assimilation to the follouint;, less sonorous environment of pause. Cf. also sectioll6below

5.2.3· Finally, there are at least some lancua~es in 11h1ch (S)FD has a very different function from (SJFW. Thus, Spanish shows neutralization of the voiceless and voiced obstruents (such as E and b/~) to voiceless obstruents (such as£) before obstruents in emphatic or careful speech, but to voiced or la~ ones(~ or~) in more normal speech, cf. Navarro Tomas 1918.61-79,107-10 and Alarcos 1961 179-9.19 Uhat is interesting in this respect is that also in my native dialect of German, SFD occurs more frequently in emphatic, careful speech than in normal speech where instead there is a shift in the syllable boundary; cf. normal Redner [te $dneG]vs. emphatic or careful [te t$na~] ' speaker' This £latter will be further pursued in section 7 beloir. 6 Tuo tendencies. "lFD and SF\'/?

6.1. If FD and Fw are different processes, t~e question arises as to ~1hether there is a general difference in function and/or domain bet\,een the t1,o processes One possible hypothesis '1h1ch might suggest; itself is tnat they differ in their respective do11ains, 1Jlth F 1:l having as its domain the syllable, and FD, tl1e \Jord ~his mignt be considered to be supported by tne fact that (int;ernal) SFD seems to occur rarely I an mmre of only t;i,10 languages in which :process occurs, namely Spanish and German, and 1Jhat is interestin;; 238 1975 MID-AMERICA LINGUISTICS CONFERENCE

about the SFD of these laneuages is that it occurs under very special, 11 emphatic 11 or 11 careful 11 conditions, cf. 5.2 3 above and 7.1.5 and 7.4.4.1 below. On the other hand, though WF~'/ is by no means rare, it does not seem to occur as frequently as WFD. (The reason that it may occur at all would, of course be due to the fact that word boundary often (especially1 prepausally) coincides with syllable boundary.) 6.2 Houever, this hypothesis probably is too facile and does not entirely agree with the facts. For there are languages l:i.ke Elean Greek ( w:i. th E. > I;/__ /f.) in uhich FW applies only word-finally. Cf. similarly Skt. E. > Q (> y/w) I _JI:., but not /_$. If the natural domain of FW were indeed the syllable, and never the word, then we would expect it to occur in all syllable-final env:i.ronments, in all lan- gµages in mrfch :i.t occurs. === 6.3. A more plausible hypothesis would therefore be the following. In section 5 .2 .2 above it has been shovm that FW can be looked upon as an assm:i.latory process (except for loss), namely as a ljg assim:i.lation (:i.n terms of sonority and/or aperture to preceding segments. Simil- arly, it can be argued that (prepausal) FD is an anticipatorl process, not (of course) to 'silence', but 00 the res~ posit:i.on of the vocal cords \iliich, even if not :i.dentical, is more similar to the:i.r devo:i.cing position than to their voicing posit:i.on. This uould explain why SFD is so rare, for unlike pause, syllable boundary is not a phonet:i.c environment and thus could not condition any (anticipatory) assimilatory change.-- If this vieu is correct, those rare instances wnere SFD is found '1ould have to be considered i,he result of the secondary generaliza0:i.on of prepausal devoicing not, as nappens usually, to word-final position, but to syllable-final environment.

~t the same t:i.me, this nypothesis might be ta~en to explain why \'TFW is comparatively rare. For ,1hile in FINAL WEAKENING AND RELATED PHENOMENA 239

prepausnl position, both anticipatory assimilation (i.e devoicing) and lag assimilation (i.e. increase in sonority and/or aperture) may occur a priori, the fact that anticipatory assimilation is generallj more conmon than lag assimilation would, in this more specific en- virollil'J.ent, provide for a greater incidence of devoicing than of weakening. 6 .4: Even this hypothesis, ho\1ever, cannot be considered entirely satisfactory. For like the hypothesis of 6.1 above, it fails to account for the fact that some languages have \Ti'·Jl without naving a general process of SF\J. Fioreover, any theory of F\J ope.eating uith the concept of (sonority/aperture) assimilation 1.1111 fail to plausibly account for the 11 ultimate 11 weakening, namely final loss To account for this development •1ould seem to require a different, more functionally oriented theory. (This is not to say, hm1ever, that on the phonetic plane assimilation may not be prominently in- volved in maJority of weakening processes.) 7. Syllabication, final v1eakening, and gemination

It seems t~at such a theory can be found by looking more carefully at now FW affects syllable structure and hm; the way it affec-r;s syllable structure correlates m. th the behavior of other syllable structure changes involving consonants.

1fu.ile there is no general consensus amonc linguists on the phonetic criteria 11hich define syllables and especially their boundaries (cf. the discussion in 3 3 3 2 above and note also the very agnostic position taKen by Kim 1971.66-76), tnere is a 5ood deal of evidence such as FJ and FD which strongly indicates the need the concept of syllable in phonological discussions, cf especially Hooper 1972 rioreover, some of the difficulties so far encountered in defining the syllable may be due to the fact that some languages (such as English) I'-ave very poor phonetic 240 1975 MID-AMERICA LINGUISTICS CONFERENCE

correlates of syllabication, even in fairly normal speech In these languages, syllables may appear only in very careful, slow, almost unnatural speech. On the other hand, however, in many languages such difficulties arise only in fairly informal, allegro speech, while in normal speech, syllabic divisions are quite clear, both to the speaker and to the hearer. Even if also in these cases, phoneticians may not always be certain about the acoustic clues and articulatory bases for syllabic di- vision, this should not prevent us from proceeding with our analysis, on the basis of the intuitive knm1ledge of both the hearer and the speaker. Finally, some of the difficulties '''hich so far have been encountered in defining the (scope of the) syllable may be the result not so much of a E!;!leral inability to do so, but rather of an inability to do so in a fashion which is generally applicable to all lanGuages. This perhaps very subtle distinction \Vlll, I hope, become clear from the subsequent discussion which shm!S that different languages~ at different times (and sometimes even at the same time) may syllabicate identical sequences in very different ways. I further hope that the subsequent discussion i:,nll show that these differences (by and large) are not random and unprin- cipled, but that they follow a very definite, hi~rarchi cal pattern.

7.1 A hieEarchy of syllabication. >'ilnle the determination of the hierarchy of syllabication, accounting for all possiole segoent sequences, 1rould be both beyond the scope of the present paper and, I believe, irrelevant for the present argument, it is possible to establish a hierarchy for the syllabication of V [+ stop] [- syll]~V sequenceso20,21 1 2 1$2 $12 [+ stop] [+ stop] [+ fric] [+ nas ] [+ llqu J y,w I l FINAL WEAKENING AND RELATED PHENOMENA 241

I.e. [+ stop] [+ stop] is mos"G likely to be separ- ated by syllable boundary, ~·.rhile [+ stop] [+ glide] is most likely to be tautosyllabic, with syllable boundary precedint; the entire cluster Compare the follo1 11ng evidence. 7.1.1 (cf Allen 1953 82 and 85) The syllabication recognized by the maJority of the Sanskrit grammarians, as \lell as indicated by the evidence o.f metrical conventions, is 1$2 .for the uhole hierarchy (and .for all CC clusters). According to one phonetic treatise, houever, the Rik Pratisakhya, both 1$2 and $12 are possible alternat- ives for the entire hierarchy (and for all other CC clusters). The latter syllabication .finds support in the structure o.f the essentially syllabic \JrJ.. ting sys- tems of India, such as the devanagari To the extent that this is graphemically possible, all consonant let- ters in a cluster are combined into one symbol, together with the necessary vm1el modifications. Thus kartsnya vill be written

Finally, the Taittiriya Pratisakhya gives the following syllabications. [+ stop] $ [+ stop] $ [+ stop] [+ fric.] [+ stop] $ [+ nas.J $ [+ S"GOp] [+ liqu J $ [+ [+ glide]

This syllabication is some 1 1hat unusual and does seem to require a few comments L~S adumbraued in 3 3 3 3 above, the nasals are somer.1hat 'uncoTYJ.fortable' in nier- archies of this sort, since on 1Jhe one hand they are stops (Just like the oral stops), but on tne other hand ~hey do have a relatively high degree of sonority due lio their nasal resonance While the usual nattern in the syllable structure hierarchies discussed in tl1is 242 1975 MID-AMERICA LINGUISTICS CONFERENCE

section seems to focus on this latter fact, a minority pattern apparently focuses on the fact that nasals are stops and, accordingly, treats nasals and stops alike; cf. similarly in section 7 2.5 below--significantly again in Indo-Iranian. -,, 7.1.2. Greek (cf. Schwyzer 1939 237 w. references) The earlies, Homeric and Lesbian, pattern of syllabication, suggested by metrical evidence, is 1$2 for the whole hierarchy (except for the fact that ~ and w do not occur after consonants). -

In later, Attic-Ionic Greek, however, [+ stop] [+ liqu.] occurs metrically as $12, first apparently in the case of £+ stop] £, later also in the case of [+ stop] l~ 2 For even later periods, there is metrical evidence that tautosyllabic CN clusters uere becoming (optionally) acceptable, at least in the case of !}! and E!!!.• 7.1.3 Latin (cf. Hale and Buck 1903.6,14 and Sch •.iyzer 1939 23'/) The metrical evidence of Old Latin indicates a syllabication 1$2 for the entire hierarchy, except for~ ',Jluc.o. has $12 (no doubt due to the fact that this sequence usually goes back to the Proto-Inda-European single segment *kU). Note that there are no other oc- currences of ~ and ~ in the environment [- syll] ~- [+ syll. J. In later Latin, $12 is optionally possible also for[+ stop] [+ liqu.]. 7 1.4 Spam.sh (cf. Alarcos 1961 178-9 and passim) The syllabication is 1$2 from [+ stop][+stop] throuGh [+stop][+ nas.J. It is $12 for [+stop][+liqu ] (except for tl, dl, cf. fn 22 above) and for stop plus glide ~ ~ FINAL WEAKENING AND RELATED PHENOMENA 243

7.1.5: German (my dialect) The syllabication always is 1$2 in the case of [+stop] [+stop]. It is 1$2 in[+ stop] [+ fricJ through [+ stop] [+ liqu.] (a) in co~pounds 1 as in Ab- fall [?a..r.")$fal], and (b) elseuhere in emphatic or very careful speech, as in A?fe~ [?ap$fal] vs. more normal [?a$pf:f-], otheri11se it is 12. In [+ stop] y the syl- labation is 1$2 in compounds (as in ab-Jaren-[?ap$ya $ g~J), and $12 else\zhere in normal speech as in Akazie [?a$ka·$tsya]); in emphatic or careful speech$ [T stop] I $yV occurs instead of norr:i.al $ [+ stop] y V (as in [?a$ka.$tsI$ye]). (\lords with ware extremely rare and are basically[+ foreign].) - 2 7 .2: A hierarchy of syllable-final \1eak:ening 3

\ The evidence given in 7.2.1-5 suggests a hierarchy of the .:ta..10\rinc; sort.

[+ stop] [+ stop] [+ fric.] 1 [+ nas.] [+ hqu.] y' \1 I.e. SF' J of stops is most likely to occur before stops, and least likely before glides 7.2.1 Late Classical/Byzantine Greek (cf. Sch

[+stop] through [+nasal]. It also occurs in t,d before 1, i.e. in the environment in which Spanisn inserts a syliable boundary between stop and liquid. 7 2.4 Celtic (cf. Pedersen 1909 430, Thurneysen 1961 74,135,136, and especially 139 and 140) SF\J occurs before [+ stop] through [+ liquid], except that *dr normally develops into geminate *ddr 7.2.5. Iranian (cf Reichelt·l909 38 and passim) In voiceless stops, SF1J occurs throughout the hierarchy In voiced stops it is found only before [+ fric~ at least in GathaAvestan. Regarding this situation, the following two comments need to be made. (a) 11 SF11/ 11 of voiceless stops occurs not only in genuine syllable-final positions, but also after #. Houever, note that similarly the MW of ( 1,/estern) Romance seems to have originally applied across #, and still does in some dialects, although in many dialects it has been subJected to various levelings, cf Lausberg 1967 passim. Similarly, Celtic 'lenition' originally applied across vord boundary, and continues to do so in (most o~ the Celtic languages, albeit in a morphologized fashion It is thus not unlikely that also in pre-Iranian, SFW took place across #, and that its historical occurrence even after## is due to similar levelings.

(b) The fact that SF~J of voiced stop~ occurs only before fnction continuants (namely before z) is reminiscent of the fact that also in one variety of Sanskrit syllabication, [+ stop] and [+ nasal] act the same, thus disrupting the normel hierarchy, cf. 7 1 1 above. 7.3 A hierarchy of gemination23 Note that this discussion is concerned only i.n th the semination frequently encountered in the first mem- ber of the clusters under discussion. Other Beminates, especially those involving assimilations (as uell as 11 spontaneous" or 11 expressive 11 geminations ), may follow completely different patterns. FINAL WEAKENING AND RELATED PHENOMENA 245

The following hierarchy seems to be suggested by the evidence discussed in 7.3.1-5, even if the evidence is not quite as straightforward, and the pattern as 11 "neat , as in 7.1 and 7.2. [+ stop] [+ stop] [+ fric.] [+ nas.] [+ hqu J y,w I.e stops are most likely to be geninatedl before y and .:!'.!' and least likely before stop 7 3 1 Old Norse (cf Krahe 1966 104-6) Gemination is found only before y and 11 uore- over, the only stops which are geminated in this environment are ~ and E_, i.e. the stops homorganic '11th z and !?. • 7 3 2 Old Irish (cf. Thurneysen 1961 74) Gemination applies only before r. ,\nd only the nomorganic stop ~ is geninated in th1s environment 7.3.3 Italian (cf. Grandgent 1927 99-108, Laus- berg 1967 46-66) Almost all consonants (except s, r, n, and 1) geminate before ;t_ (including y < 1). Hote °fha"t ~seems to have under~onc degemination tor SFW), becoming It s (vs. !El> z). Consonants also generally geminate before nsecoooary;r ~'i.e. before~<~ IC __ V.

Be~~re r, the only remaining liquid in inherited \lOrds, only :§: (and .f) geminate, other consonants do not 7 .3 .4. ''lest Germanic (cf. Krahe 1966 104-6, Braune- Mi tzka 1959 94-6 with 104, 149) All consonants (except r) geminate before y Be- fore ~' gemination is found only in the (homorganic) velars. Before liquids, only E' !, k are found to geminate. Gemination is rare before nasaTs, but does seeffi to 246 1975 MID-AMERICA LINGUISTICS CONFERENCE

occur in £, !, ~' and ~o Sanskrit (cf Wackernagel 1896·112-4-) Gemination may occur throughout the hierarchy, but various authorities deny this, or limit or extend it in its scope. 7.4- Interpretation. 7.4-.1 As far as syllabication is concerned, it is noteworthy that the available evidence strongly sup- ports the assumption of 1$2 throughout the hierarchy as the original pattern of (Proto-) Indo-European. It is above all tne dynamic evidence of Greek and Latin which suggests that this pattern was later subJect to a syllable boundary shift (SBS) \mrking its way 11 up 11 through the hierarchy The most radical instance of SBS would seem to be found in the variety of Sanskrit which permits $12 throughout the hierarchy 7 .4 2 Concerning SF\'r, it is remarkable how close- ly the pattern of this process agrees with that of syllabication and SBS. Notice specifically the tight fit between Span 1$2 and SFW, down to the fact that tl, 2:1 which exceptionally syllabicate as 1$2, also exceptionally undergo SFw. Compare further the striking correlation between the syllabication patterns of Sanskrit and tne SFW patterns of Iranian, i.e. of the t1.vo maJOr bran- ches of Indo-Iranian. 7.4 3 What is remarkable and seems to be in need of an explanation is the fact that also the pattern of 11 11 gemination, though much less neat , seems to correlate significantly with the patterns of syllabication (and SBS) and of SFWo Thus, Celtic gemination begins precisely in the environment where SFW leaves off, namely before r. This seems to be true also for Italian, to the extent that it had SF\l, cf. Lat supra • It~ sovra (beside soura) vs Lat fabru- • It. fabbro. Moreover, this seems to correlate with the fact that the environment before liquid is the only environment in which Lat1n 1 the ancestor language, had SBS from 1$2 to ~12. FINAL WEAKENING AND RELATED PHENOMENA 247

Finally, note that Sanskrit, vn~ich offers 1$2 and $12 in the entire hierarchy, is also the only languae;e 1.ri th possible gemination throughout the hierarchy 7.4.3.1. The earlier views of de Saussure (1889) and Fouche (1927 81-91) do not seem to account for this correlation.

According to de Saussure, it is only before z,~,£, l,m,n that gemination should be possible, because in that-environment there can be no distinction between single and geminate (dental) stops, both being realized as 'implosive' stops followed by 'explosive' stops in the next syllable (at least after short vowels). In other environments, geminates are said not to be possible. Considering the evidence of Sanskrit in favor of gemination also before obstruents, and considering the Italian (pattern) contrast between sovra/sopra and fabbro, neither Qf de Saussure's claims can be considered acceptable.~4 7.4.3 2· According to Fouche, gemination before r,l,z,w is the result of an 'articulatory difficulty' Tu1tli clusters of this sort]. This may be alleviated through simplification (loss), prothesis, anaptyxis, metathesis, or syllabificawion of the second me~ber of the cluster. But in order to preserve the 'explosive' stop of the clusters, the language may 'reinforce the articulation of the former' by increasing its duration. (181-91, a completely different and hiGhly doubtful explanation is given ibid 46-60.) Again, also Fouche's analysis suffers from being incomplete, by not considering the fact that gemination is possible also "higher up" in the hierarchy Ho rever, some of Fouche's ideas vill turn out wo be quite a pro- pos. 7.4.3.3 One interpretation which might readily suggest itself can easily be dismissed, namely an interpretation focusing on the fact that the observed pattern of gemination might be considered the result of the fact that in many languages one can find 11 aP'.lbisyllabic 11 stops, 248 1975 MID-AMERICA LINGUISTICS CONFERENCE

such that syllable boundary seems to occur somewhere in between the closure and the release of the stop, especially in the lo\.ler spectrum of tm hierarchy. Such ambi- syllabic stops can be ob~erved in many varieties of German, as in rattle [ra Stle]. Hm.rever, if such ambi- syllabic stops really were the source of gemination, then one would expect gemination of orii;inally single seGments to occur most frequently in intervocalic environments 1.1here ambisyllabici tytis most readily found, as in GerPl rette [r£$te J or [r£ Sta J. Ho\mver, this is precisely""""t'fi'eenvironment 1.1here such gePlination is found most rarely 7 .4 .4 A more likely e:A'1Jlanation \muld seem to be possible within a more general theoretical framework 11h1ch considers the functional relationship of SFW and SBS to each other, to the syllabication pattern 1$2, and to t;emination 7.4 4.1 The great similarity in hierarchical patlarning between SFW and $12 is suggestive of a special relationship This impression is reinforced by the fact that where Spanish opposes normal SF,/ to emphatic/careful SFD, Germ3n opposes normal $12 to emphatic/careful SFD (except for [+ stop] [+ stop] and compounds) In ad- ~1t1on, note that historically, $12 seems to be secondary, coming about as the result of SES. If we assume that it is this process of SES (rather than tne uattern $12) wnich is related to the process of SF1l, i:;'IJ.e functional relationship beti,1een the two phenome- ~a becomes clear Both SBS and SF" (at least, if earned through to its ultimate conclusion, namely loss) lead to ouen-syl- bbl~ structures I e , both changes (ultimately) produce the pattern which sePms to be most favored on the spea~er's end of the speech process (given the prior existence of 1$2 patterns) Note in this regard that both processes figure prominently in the Slavic open- syllable conspiracy, as in *sed$lo 'seat' > se$lo, se~lo, or seSdlo depending on the dialect. ~~- FINAL WEAKENING ANO RELATED PHENOMENA 249

Notice additionally that this interpretation permits a principled explanation for the problem discussed in 6.2 and 6.4 above, namely the question as to why some languages have WFU \11th.out also having a general SFv-J. For while internally, SBS is in many cases a viable alternative to SFw in producing open syllables, in abso- lute final, i.e. prepausal environment, there is no sacn alternative--only WFN can produce open-syllable structures in this position. 7 .4 .4.2. On the other hand, howeva-; in producing open-syllable structures, th~ two processes either weaken or delete underlying stops (and other consonants) or, as a result of SBS, shift them into an environment in \1hich they are less clearly perceptible. For the hearer, it is rather the structure 1$2 \ihich provides the optimal acoustic clues concerning the identity of these underlying segments. 7 4.4.3. This functional interpretation of the changes SFW and SBS and of their difference fro111 the pattern 1$2 now permits a principled explanation of ge111ination and its pattern similarity to $12 (and SF~'1), an ex~lanation reminiscent of, but not identical to Fouche's Gemination may be conceived of as a reaction to SBS (from 1$2 to $12) and its effects on vne nearer, namely as a phonological 'blending' betueen the speaker- favored pattern $12 and the hearer-favored original pattern 1$2, yielding a conpromise pattern 1$12. This i.1ould most readily explain vhy gemination occurs most commonly in those clusters \vhicn also most commonly have (SBS to) $12 (and so forth, uit.r.. decreasing likelihood, 'up' through the hierarchy). It would also explain wny the pattern of gemination is so much less 11 11eat 11 than the other patterns observed As a scondary, analogical process, it need not be expected to be regular. 250 1975 MID-AMERICA LINGUISTICS CONFERENCE

8 Conclusion '.lhile final weakening may, at least in most of its staGes, be considered (lag)---a8similation, its ultimate function see~s to be comparable to that of syllable boundary shift, namely to bring about (s~eaker-favored) open-syllable structures On the other hand, final devoicing seens to be more clearly assimilatory, although at least in some languages it may, apparently secondarily, be used with a particular function, namely as an emphatic or careful alternative to either final \Jeakening or syllable bound- dry s~ift. Considering that syllable-final devoicing seems to be exceedingly rare and considering its very special function in the tuo languages where it has been found, it may perhaps not be idle to speculate that it is only under such special, functional conditions that devOIC:lng may have the syllable, rather than the word, as its domain. Finally, it appears likely that at least one process of gemination may be the result of analogi• al, rather than primary, purely phonetic developmenL 1 namely of the blending of an older, hearer-favored 1$2 pattern with an innovated, speaker-favored $12 pattern.

NOTES lResearch on this paper was in part supported by a 1975/76 grant from the University of Illinois Research Board 2Especially interesting is Stampe's claim that even languages uith exclusively CVCV structure have word- final devoicing, 'but without overt manifestation. Tnis claim appears to be supported by the pronunciation, in such languages, of foreign uords witn final voiced ob- s"Gruents, uhich, if they are pronounced at all, are c~aracteristically devoiced '(445) Unfortunately, Stampe does not provLde any examples for this developme:t. FINAL WEAKENING AND RELATED PHENOMENA 251

3Throughout this paper, the symbol $ will be used to indicate syllable boundary. 4cf. the fact that [-tense] obstruents become [-voice], but remain [- tense] in the environment r: ~g~~~J $ ~-' as in hat sie [hat$~i J vs. nimm sie [nim$zi ] • 5The fact that insights of this sort--Hhether they may ultimatdy turn out to be factually correct or not-- can be found in traditional, neogrammarian writings should lay to rest any notions that such fairly abstrac~ analogical developments as rule ~eneralization cannot be captured by traditional historical lin;;uistics, but only by generative phonological theory. Evidently, they could. (Uhat is true, however, is that they could not be captured as easily or readily ) 6Note that Brugmann (ibid ) mentions the fact that there are also German dialects (Seifhennersdorf, Soest) .1hich show a pattern sir:11lar to that of Sanskrit 7Even Szemereny1 (1973.70) finds Andersen's explanation 'surprising.' 8No such neutralization is found, however, in final fricetives This is interesting in light of the observ- ation of .\.11dersen ( 1972 16-8 u fn 9) that in East Sor- bian final devoicing 'the segments \Jhich are unmarked nith resnect to the feature [+ continuous] /d b/, be- coP1e unmarked for voicing [ chronolot;icallyj earlier than the marked segments, /z z/ 1 \I/hat this COJ11bined evidence sugGests is that, as Andersen already suspect- ed, there may be a hierarchy of susceptability ~o FD. 9For reasons of space I vill in t_1e follo llng re- frain from giving individual examples for the various C11anges, but rather refer the reader to 'JOrks in \ 'hlCh such_examples can oe found--except uhere I have not been able to find such sources or 1,~ere tne facts are too uell knoFn to require such reference.. I hope that this listing inll shou how generally known these phenomena are in traditional historical literature. (1fote that 252 1975 MID-AMERICA LINGUISTICS CONFERENCE

this listing is by no means complete.) lOThese two changes apparently were chronologically separate from each other; cf. Gauthiot 1913 J02-6 uith earlier references. llFo~ environmental restrictions beyond (i.e following) the syllable boundary, cf. 7.2 below. 12According to Allen ( 1962 98) this lhddle Indo- Aryan loss is a 'complete assimilation to silence, as stops before another stop are completely assinilated to it.' In light of the fact 1Jha-r; similar ueakening losses may occur also internally, in syllable-final position ( 1vhere no 'silence' follows), this explanation cannot be considered acceptable. 13Vennemann offers the interesting argument that t.ne peculiar vOJCing sandhi of Sanskrit is due to the fact that (i) uord-final oral stops may become unreleased, unaspirated, and ueakened, and that (ii) lexic- al representations, in his opinion, consist of the 'pronunciation in isolation' forms of given uords. It is this ueakness, then, and not uord ooundary, which in 111s v1eu accounts for the fact tnat word-final stops follou a different voicing sandhi rule from the 'normal' internal stops.-- Considering the alternative explan- ations discussed earlier in this paper, further, inde- pendent evidence uould seem to be required to make credible Vennemann's hypothesis that underlying forms are identical \li th '"pronunciation in isolation11 forms'. 14There is one maJor class of exceptions, namely dissimilatory developments of tne type OHG [oxso] > :r.c1G [oksa J. ~Iouever, as Oicel. einn > Nicel. [ei

15This is not to say that nasals may not be weak- ened. Houever, in such \leakenings (such as [m] > [,ii] > [OJ or [n] > [n]) relevant processes seem to be parallel to the above oral) hierarchy. It is true, they may eventually merge into this oral hierarchy, by denasaliztion (as in [w] > [w]). However, the converse change, from oral to nasal hierarchy, does not seem to occur in final \Jeakening processes.

161fotice in this respect the apt remarks in de Saussure 1916 72.

17Also Middle Persian seems to have had IHl.

18That the changes of .E. to .£ I _ r, 1, or to r..1 oefore nasal (and similarly to x before voiceless ob- s~ruents) must be very early chan~es is indicated by the fact that in other positions, PIE *o \l8S lost in all of Celtic, i e presumaoly in Proto~Celtic (An 'exception' to this loss would be forms 111.ce PIE *nenk\le 1 5 1 > Oir. c6ic, uhere p, of course, underi1ent distant assimilation to the fol!mnng l\Y prior to its general loss.) - 19Hote tnat Navarro Tomas's stawel11ent that before ~' .E. is regular even in normal speech if the preceding vowel is accented, is not necessarily incompatible with this statement--if we permit a possible equation of accented and emphatic environments. 20It may be noted that this hierarchy \lOUld seem to necessitate certain revisions in the recent attempt of Hooper (1972 533-8) to propose a urm.rersal definition of syllables Hooper proposes the following rules for syllabication. I (a) ¢ --> $ I [+ syll ] [- syll ] [+ syll ] ( b) ¢ --> $ I [ + syll.] [ -syll] 0-[- son.] [~~~~ ] [ +syll.] In addition, sne proposes a set of rules permitting possible exceptions (1) ¢ --> $ I ~~~·] ~- [+ lat J hence d$1, 254 1975 MID-AMERICA LINGUISTICS CONFERENCE

son. ] -+ cor. (11) 0 --> $ I ++ cont. [+ cons ] hence s$1,s$r ~- + voe ~- -~ l strid:. - son. 1 \hence $st, (11:0 0 --> $ I _ [! ~~~~d ~: ~~i~;] { $sp, $sk, [- voe. ] $s-1~ - voice ~- As t'IJ.e hierarchy in 7 1 shows, the number of 'possiole exception' rules must be consideraoly increased, as well as ranked in terms of the degree of 1 poss1b1l1ty'. 2lrt should come as no surprise that the nierarchy (of the second members) in the clusters here examined is very similar to tne earlier sonority or aperture hierarchies discussed in 3.3.3.2 above. Sinilar hierarchies have been proposed also by Fouche (1927 3-14) and Hooper (1972 533-8 and 1973 70-102). 'fnat is especially interesting in this respect is the fact that similar hierarchies have been found to be supnorted also by other evidence i.1h1ch is no-c ooviously- connected 11th the syllable structure phenomena here discussed Cormare Z1 11cky 1972 for a hierarchy governing various pnenomena in En8lish allegro speech, such as ~-syncope and otner a-deletions, r-syllab1f1cat1on, and (stop) enen"Gb.esis- Si:r:nlarly, -Hankamer and Aissen ( 1974 ui th earlier references) argue for the need for sonority ~ncrarclnes to account for ?811 and Hungarian assimilations The fact that tnes:;hierachies nave turned out to be useful, as 1 1ell as necessary, in so many differ- en~ Gheoretical and prag~atic contexts, seems to strongly suggest that tnere must be some 11nGu1st1c validity to them 22This chronological difference may be the result of a 'solidarity' of all [+ stop] l:_ clusters uith the ~l, dl clusters, \fuich are universally disfavored (even if not impossible) in tautosyllabic environment. (Cf. also Hooper's rule (i) in fn. 20 above.) FINAL WEAKENING AND RELATED PHENOMENA 255

23cf. the introductory remarks in 7.1 above. 24de Saussure's argument concernin5 the Proto-Indo- European merger of tt and t in this environment, :10\T- ever, appears reasonable enough Given a syllabication 1$2, it is indeed conceivable that /met-tro-/ and a sequence like /met-ro/ would have been realized identic- ally as met$ro. However, such a neutralization evidently need not be assumed in languages with $12, uhere trie t1vo could easily be distinsuished as met$tro vs me$tro.

BIBLIOGRAPHY Alarcos Llorach, Emilio. 1961. Fonologia espanolo. 3rd rev. ed. Madrid Editorial Gredos. Allen, U. Sidney. 1953. Phonetics in ancient India. London Oxford University Press. 1962. Sandhi. The Hague and Paris Mouton. Andersen, Henning. 1969 (a). Inda-European voicing sandhi in Ukrainian. Scando-Slavica 16 157-69. -----. 1969 (b). The change of *ot to *od a central Slavic innovation. Die \lelt der Slaven 14.315-30. 1972. Diphthongization. Language 48.11-50. Anttila, Raimo. 1972. An introduction to historical and comparative linguistics. Ne\'l York i1acMillan. Arnholtz, A., and C. A. Reinhold. 1936 Einfuhrung in das danische Lautsystem. Leipzig Harrassowitz. Bloch, Jules. 1934. L'indo-aryen au v~da aux temps modernes. Paris .A.drien-1'.iaisonneuve Braune-Mi tzka == ,1ilhelm Braune. 1959. Althochdeutsche 1 11 Gramnatik. 9th ' Auflage , revised by 'Jalther I11tz- ka. Tubingen Niemeyer. Brugmann, Karl. 1897. Grundriss der vergleicnenden Graomati~ der indogermanischen Sprachen, 1 Strass- ourg Trubner. Buck, Carl Darlin. 1904. A grammar of Oscan and Umbri- an. Boston. Ginn. 256 1975 MID-AMERICA LINGUISTICS CONFERENCE

Chen, I1atthew Y. 1973 Cross-dialectal comparison a case study and some theoretical considerations. Journal of Chinese Linguistics 1.38-63. [Non vidi.] Diver, ./illiam 1953. Tne relation of Lai:;in and Osco- Umbrian Columbia University Ph.D dissertation [Non vidi ] Ewert, Alfred 1943. Tne French language. 2nd ed London Faber 8 Faber. Fouche, Pierre 1927. Etudes de phonetique g~n~rale. Paris Les belles lettres Gauthiot, Robert. 1913, La fin du mot en indo-europ~en. Paris Geuthner Grammont, f iaurice. 1939. Trai te de phonetique. 2nd revised ed Paris Delagrave. ' Grand3ent, Charles H. 1927. From Latin to Italian. Cambridge Harvard University Press. Eale, 1 hlliaM Gardner, and Carl Darling Buck. 1903. A Latin grammar. Reissued 1966, University, Ala- oama University of Alabama Press. :=an~amer, Jorge, and Judith Aissen. 1974. The sonority hierarchy. Papers from the Parasession on Natural Pnonology, 131-45. Chicago Chicago Linguistic Society. :dock, Hans Eenrich (In press.) Revieu of Anttila 1972 To ap~ear in Language. 2ooner, Joan B. 1972 The syllable in phonological · theory LanGua~e 48.525-40. 1973. Aspects of generative phonology. UCLA Ph D dissertai:;ion. Reissued 1974, Bloomington Indiana University Linguistics Club. [References are to the 1974 reissue.] iync;n, Larry Il 1975 Phonology i:;heory and analysis ;.-eu Yor....c et alibi ~Iolt, Rinehart and '!inston. Izzo, Herbert J Tuscan and Etruscan. Toronto Uni- versity of Toronto Press. Jespersen, Otto 1904 Lehrbuch der Phonei:;ik ,11.l'"unor- FINAL WEAKENING AND RELATED PHENOMENA 257

ized vransl. from the Danish by H. Davidsen Leipzig and Berlin Teubner. Kim, Chin-·Ju. 1971. Experimental phonetics A sur- vey of linguistic science, 16-128, ed. by •l. O. Ding\lall. Colle~ Par!<:, MD Linguistics Program, University of Maryland. Krahe, Hans. 1966. Germanische Sprachuissenschaft, 1. ' ,sammlung Goschen, 238.) Berlin de G~uyter. Lausberg, Heinrich. 1967. Romanische Sprachuissen- schaft, 2. 2nd ed. (Sammlung Goschen, 250.) Berlin • de Gruyter. Malmberg, Bertil. 1948. La structure syllabique de l'espagnol. BoJftim de filologia 9.9g-102. Re- printed in Bertil Malmberg 1971.389-404. [Ref- erences are made to the pagination of Malmberg 1971. J • 1952. Occlusion et spirance dans le systeme consonantique de 1 1 espagnol r1elanges de philolo- gie romane offerts a Karl Inchaelsson, 356-65 Goteborg. Reprinted in I1almbe:ig 1971 405-13 [References are to the paGination of the reprint J 1963. Phonetics. rew York Dover. • 1971. Phonetique generale et roMane The gaGue Nouvon r1artinet,, _"l.ndre. 1970. Economie des changements phonetiques. Berne Francke. Inller, D Gary. 1973. On the motivation of phonological change. Issues in linguistics papers in honor of Henry and Renee Kahane, 686-718, ed by B B Kachru, et al. Urbana, Chicago, London University of Illinois Press Ihranda, Rocky V. 1974. Sound change and otner phonological change. innnesota 1'1orking Papers in Lin- guistics and Philosophy of Language, 19?4 2 49-81 Navarro Tom~s, Tom~s. 1918. IIanual de urononciacion espanola. IIadrid Centro de Estudi~s Hist6ricos Pedersen, Bolger. 1909. Vergleichende Grammatik der keltischen Sprachen, 1. Gottingen V2ndenhoeck 258 1975 MID-AMERICA LINGUISTICS CONFERENCE

und Ruprecht von Planta, Robert 1892 Grarrimatik der oskisch-um- brischen Dialekte Strassburg Trubner Postal, Paul fvl. 1968 Aspects of phonological theory Ne1 r York Harper &- Rou •

.flulgram, Ernst 1970 Syllable, word, ne}..'llS, cursus 0 The Hague and Paris Ilouton. Reichelt, Hans. 1909. Awestisches Elementarbuch. Eeidelberg Hinter Safareuicz, Jan 1969 Historische lateinische Gram- matil\:. Lalle VE3 lhemeyer. de Saussure, Ferdinand. 1889. Sur un point de la phonetique des consonnes en indo-europeen flSL 6 246-57. • 1916 Cours de linguistique generale 4th, slightly revised reissue, 1949. Paris Payot. Schane, Sanford. 1972. Natural rules in phonology. Linguistic change and generative theory, 198-229, ed. by R. P. Stock\vell and R.K.S. l1>iacaulay. Bloo- min~ton and London: Indiana University Press Sch\,ryzer, Eduard 1939. Griechische Grammatik, 1. P.iunchen Beck Stampe, David 1969 The acquisition of phonetic representation. Papers frorri the 5th Regional !leeting of tne Chicago Linguistic Society, 443-54. Jzerrierenyi, Os\1ald 1973 11arked-hllP1arked and a problen of Latin diachrony. Transactions of the Philologic- al Society 1973 54-73

Tirnrneysen, Rudolf 1961. _1_ srammar of Old Irish. Re- vised and enlarged ed., transl by D. A. Binchy and O. nergin Dublin The Dublin Institute for :.dvanced Studies Voillant, Andre. 1950 Grammaire corriparee des langues slaves, 1 Lyon IAC. Vennemann, Theo 1968. U-errrian phonology. UCLA Ph.D. dissertation [i'Ton vidi.] FINAL WEAKENING AND RELATED PHENOMENA 259

Vennemann, Theo. 1972. Sound change and markedness theory. on the history of the German consonant system. Linguistic change and generative theory, 230-74-, ed. by R.P. Stockwell and R.K.S. Macaulay. Bloomington and London. Indiana University Press. 1974-. Words and syllables in natural generative grammar. Papers from the Parasession on Natural Phonology, 34-6-74-. Chicago. Chicago Linguistic Society. \/ackernagel, Jakob. 1896. Altindische Grammatik, 1. Gottingen: Vandenhoeck & Ruprecht. Zwicky, Arnold M. 1972. Note on a phonological nierarchy in English. Linguistic change and generative theory, 275-301, ed. by R.P. Stockwell and R K s. Macaulay. Bloomington and London Indiana Univer- sity Press.